Paul Graham: skip entrepreneurship courses, give founders free time
Paul Graham, writing on paulgraham.com in an essay dated August 2026, argues that universities are already well equipped to prepare startup founders and do not need a new "entrepreneurship" curriculum. He frames Y Combinator as the natural authority on the question: YC has had 20 years to refine its model of what a promising founder looks like, and the trait its partners screen for is simple, being good at building things and having a habit of doing it. That skill, he writes, comes from studying computer science, mechanical engineering, molecular biology or similar subjects, not management or finance, so the right preparation is for universities to keep teaching what they already teach well. He allows that building can be broadly defined: he credits Steve Jobs's study of calligraphy as one reason Apple came to dominate desktop publishing, and notes Mark Zuckerberg was a psychology major, not a computer science major.
Graham says only two real changes are needed. The first is convincing students that starting a startup is a viable option at all. He points to YC's own application data as a proxy for interest: Harvard alumni apply to YC at about twice the rate of Yale and Princeton alumni, a gap he attributes to Harvard's stronger startup culture rather than to any difference in student ability. The most effective fix, he argues, is exposing students to founders in real life, especially ones who talk candidly about their early, clueless mistakes. He proposes a rough formula for how inspiring a founder is: their wealth and fame divided by how much older they are than the students watching. On that logic, a founder in their mid twenties, three years into a startup valued at a couple hundred million dollars, is more effective on campus than a famous billionaire, since despite being roughly a twentieth as rich and famous, they are about twenty times easier for students to identify with.
The second change is giving students more free time for their own projects, for four reasons: projects are the deepest way to learn a subject, they are how cofounders discover each other by working together, they make the self-directed nature of a startup feel familiar, and unplanned side projects are where the best startup ideas actually originate, since the people drawn to them are technological bellwethers. Graham illustrates the point with Microsoft and Meta, both of which he says got started during Harvard's reading period, the unstructured gap between the end of classes and final exams. Removing the pressure of an imminent deadline for a few weeks, he writes, "resulted in two trillion dollar companies." He expects universities to resist loosening students' schedules, partly from an instinct that achievement requires active intervention rather than simply leaving students alone, and warns that any student projects should stay unofficial rather than be formally sanctioned by the university, since official recognition tends to collide with campus rules. He cites Bill Gates, who broke Harvard rules by bringing the non-student Paul Allen into the computer lab to work on Altair Basic, and Zuckerberg, who was placed on disciplinary probation over Facemash, as examples of projects that would not have survived official oversight.
Graham closes by describing what universities should not attempt: actually teaching students how to start a startup in a classroom. Running a startup, he writes, is incompatible with being a full-time student, so any class that tried to simulate the real thing would have to either withhold funding or force students to abandon a startup that started succeeding. He singles out business plan competitions as a common but counterproductive substitute, calling them "not merely useless but positively misleading" because they teach students that pitching investors, rather than building something users want, is the essential step in starting a company.
Key facts
- Paul Graham argues universities need only two changes to prepare founders: convince students starting a startup is viable, and give them more free time for their own projects, rather than adding entrepreneurship curricula.
- Harvard alumni apply to Y Combinator at about twice the rate of Yale and Princeton alumni, a gap Graham attributes to visible startup culture rather than differences in student ability.
- Graham says Microsoft and Meta both started during Harvard's reading period, the unstructured gap before final exams, and that removing deadline pressure for a few weeks "resulted in two trillion dollar companies."
- Bill Gates and Mark Zuckerberg both broke Harvard rules over undergraduate projects, Gates by bringing non-student Paul Allen into the computer lab and Zuckerberg through the Facemash disciplinary case, which Graham uses to argue such projects should stay unofficial.
- Graham calls business plan competitions "not merely useless but positively misleading" because they teach founders that impressing investors matters more than building for users.
Why it matters
The essay pushes back on a common instinct in higher education, that fostering entrepreneurship means building a dedicated curriculum around it. Graham, drawing on Y Combinator's two decades of screening applicants, argues the opposite: the skill that actually predicts founder success is being good at building things, which existing science and engineering departments already teach. His proposed fixes, visible near-peer founders and unstructured time, are cultural and scheduling changes rather than new courses, which makes the essay a direct argument against the entrepreneurship-program trend it is implicitly responding to.
Who it affects
University administrators and department heads deciding how to support student entrepreneurship, computer science and engineering faculty, students weighing whether to start a company during or after school, and anyone designing formal "how to start a startup" coursework or business plan competitions that Graham argues do not work.
How to use it
Graham's practical recommendations for a university: put relatable founders, not just famous billionaires, in front of students; loosen scheduling so students get Harvard reading-period-style stretches without assignments due the next day; skip formal entrepreneurship classes and business plan competitions; and let student side projects stay unofficial rather than folding them into sanctioned programs, since campus rules and ambitious projects tend to collide.
How solid is it
This is a personal essay from a single author, Y Combinator's cofounder, not an empirical study. Its evidence is anecdotal, Steve Jobs, Bill Gates, Mark Zuckerberg, plus YC's own unpublished application-rate comparison between Harvard, Yale and Princeton, rather than a dataset a reader can check. The claim about Microsoft and Meta both starting during Harvard's reading period is stated without a citation or further detail. Weight it as an experienced insider's argument, not as peer-reviewed research.
Risks and caveats
The examples Graham leans on, Microsoft, Meta, Apple, are survivors chosen after the fact, and the essay does not address how many reading-period or free-time projects failed to become anything. Several figures are given only approximately or relatively: "more than 10%" as an upper bound on how many students will ever found a company, "a couple hundred million" as an example valuation, and "twice the rate" and "a twentieth" as ratios without the underlying absolute numbers. The essay also does not name Graham as its author in its own body, propose any concrete mechanism for how a university would shorten coursework or lengthen reading period, or give a numeric age gap behind the inspiration formula it describes.
“Merely eliminating it for a few weeks resulted in two trillion dollar companies.”
— Paul Graham, "How Universities Should Prepare Founders"